Precise　pupil　features　detection　is　an　important　factor　for　screening　diabetic　retinopathy.　Some　interferences　caused　by　reflections,　eyelashes　and　eyelids　in　pupil　extraction　need　to　be　solved.　This　paper　presents　an　algorithm　to　precisely　estimate　pupil　features:　pupil　center　and　pupil　radius.　The　system＇s　hardware　component　allows　for　high　frame　rate　image　acquisition　under　infrared　lighting　conditions.　We　use　several　real　time　image　processing　techniques　to　estimate　the　pupil　size　under　the　influence　of　corneal　reflection　and　eyelid　occlusion　conditions.　The　experimental　results　show　a　good　robustness　and　accuracy.